(1) Field of the Invention
The present invention relates to a phase-interpolation circuit, more particularly, a phase-interpolation circuit which can avoid short-circuit current and generate linearly distributed phase-interpolation signals.
(2) Description of the Prior Arts
Multiphase systems are widely applied in the data-recovery systems and the phase-lock loop circuits. The multiphase systems are also a main trend of design in the communication systems nowadays.
In a conventional data recovery system, after applying the equalizer to recover high frequency decayed signals resulting from noise of communication media and signal decay, the correct sampling of data streams still depends on the correct clock pulse. In addition, the rising/falling edges of the clock pulses need to be in the middle of the data period to sample them correctly. The conventional data recovery system uses a phase-lock loop circuit to achieve the clock-recovery in order to correct the received data stream and sampling clock pulses. However, several disadvantages exist when the phase-lock loop circuit is used for recovering the clock-pulse, including: (a) Longer lock time. The function of the phase-lock loop circuit rectifying the phase is to use the output signal reference frequency and the voltage-controlled oscillator to compare with each other and transfer the result to the voltage signal. The feedback of the input of the voltage-controlled oscillator rectifies the frequency in order to accelerate or slow down the phase. The process of the rectification of the phase has to compare many clock-pulse cycles with each other in order to achieve the phase needed. Maybe several hundreds of the clock-pulse cycles are needed for the comparison. As a result, longer lock-time is required. (b) Phase noise. The input voltage of the voltage-controlled oscillator of the phase-lock loop circuit will lead the frequency drift when the noise disturbs. The result above is the phase noise. When clock-pulse is recovering, comparison of phases also depends on sequence of the received data stream. Because the rising and falling edges of the signal do not vary when a long and same logical signal inputs, the phase detector won't work. Under such condition, the phase-lock loop circuit is unable to maintain the phase-lock state and the frequency will start to drift, and, as a result, the phase noise is generated in the frequency spectrum. (c) Only one receiving channel is available. Only one receiving channel can be provided when we achieve the clock-pulse recovery by using the phase-lock loop circuit. We have to recover the clock-pulse by using more phase-lock loop circuits for a plurality of receiving channels.
Therefore, multiphase systems have become a trend for data-recovery systems during the recent years. Since multi-phases may be distributed in one clock-pulse, the above mentioned phase-lock process can be achieved by selecting a suitable sampling clock-pulse. Not only phase-lock time is shorter, we can also provide a phase-lock loop circuit for use with a plurality of receiving channels. There are several methods to generate multiphase signals, including: (a) Delay-Lock loop. Using a long series of delay chain, such as two inverters connected in serial, to generate clock-pulse signals having different delay duration so as to form the multiphase signals. The advantage for this method is the stability. However, it also incurs the drawbacks of requiring lots of delay units for covering 360 degrees of clock-pulse phases, higher cost, fewer applications available, more electric power decay, and inevitable noise disturbance occurred from electrical power lines. (b) Multiphase VCO. Currently, the newly developed multiphase oscillators are able to generate refined differences of phases uniformly distributed within one clock-pulse. In addition, the number of the generated multiphase signals can be the power of two. However, one of the difficulties is to avoid the problem of the multiphase oscillator model. The circuit layout is also an important consideration for this method. (c) The phase interpolation. The phase interpolation is one of the ways to generate the multiphase signals. A middle phase can be output by inputting different phases. The way to use the phase interpolation is easier and is the trend for future technology.
The advantages of the phase interpolation in comparison with the above mentioned conventional techniques are following: (a) It won't be limited by the delay time of the delay unit to decide the distribution density of the multiphase. The multiphase density and number can be easily decided by phase interpolation. The cost and the decaying power will balance in the design. The system is also very stable. (b) There is no disadvantage of the multi-oscillating model. And the number of the phases will increase by doubling the number of the input phase after the interpolation. For example, we can interpolate 8 phases once if we need 16 phases. (8*2=16) If we need 20 phases then we can interpolate 10 phases once. (10*2=20) Or, to interpolate 5 phases twice can also obtain the same result. The design is very free. (c) Because the phase-interpolation can produce the local multiphase clock signals by using fewer phases of the globe clock signals, therefore the area of the wiring and the number of the clock-pulse buffers will be fewer than the decay-lock-loop circuit and the multiphase oscillator in the application of the multi-receiving channels.
Conventional phase-interpolation methods include the type of non-full swing signal and the type of full swing signal. The phase-interpolation of the non-full swing signal type generally employs the V-to-I current adder. The middle phase can be generated by adding two signals with two different phases. A set of binary code or temperature code is applied to control the tail current of the circuit and to rectify the weighting of the adding signals. As a result, the phase of the interpolation produced can be controlled to drift backward or forward. The distribution of the phase is not distributed in one clock-pulse period uniformly because the phase of the interpolation of the analog signal is decided by the rating of the tail current of the current adder. The switch of the phase boundary is non-seamless. We have to get the analog signal near the position of the voltage-controlled oscillator for the phase interpolation because the input clock signal is non-full swing signal. The result above will limit the application of the data-recovery system of the multi-receiving channel described above. The phase interpolation of the full swing signal will provide fewer globe clock signals of the phase to produce near local multiphase clock. The characteristic above is the advantage of the full swing signal type.
The disadvantages of the phase-interpolation method in the type of conventional full swing signal type include: (a) The decaying power of the short-circuit current is large. (b) The nonlinear phase distribution. (c) The duty cycle of the clock-pulse output is not 50%. The phase-interpolation circuit of the full swing signal type in prior art is composed of two inverters 11 and 12 as which shown in FIG. 1A. And the inverters are composed of the CMOS devices, as which illustrated in FIG. 1B. The principle of the operation of the devices is to provide two-level clock signals CK1 and CK2. As shown in FIG. 1C, the phase of CK1 is before the phase of CK2 in order to produce an output signal with middle phase by short-circuit of the two inverters. However, as indicated by the arrow shown in FIG. 1B, the circuit will produce a short-circuit current Isc when the time is T1 and T2 and when the clock signal CK1 is in the input end 111 and the clock signal CK2 is in the input end 121. The disadvantage of large decaying power in the circuit exists. And the phase of the interpolation signal produced cannot be controlled in the middle range of the phase of the original signals.
Please refer to FIG. 1D which is a schematic diagram illustrating an improved prior art circuit employing the above mentioned method. The circuit shown in FIG. 1D can correct the defects of the phase of the interpolation signal to be in the middle of the two original phases by fine tuning the current ratio of both discharging paths (e.g., to control the ratio of tail current). However, such kind of method can only improve the linearity of the distribution of the phase by means of the ratio of current (e.g., W and 1-W shown in FIG. 1D), which is difficult to be controlled concisely. This method cannot achieve the yield of the mass production and cannot maintain the duty cycle of the original clock-pulse. It will require an additional duty cycle correction circuit to maintain the 50% duty cycle.
FIG. 2A describes another prior art circuit, in which a phase-interpolation circuit that can avoid short-circuit current is provided. However, it also incurs the drawback of being only able to produce the interpolation phase for the falling edge of the clock pulse, but cannot provide the interpolation phase for both the rising and falling edges of the clock pulse (as which shown in FIG. 2B). Therefore, such kind of circuit can only be used for doubling the frequency of multiphase clock pulse. And it will not produce a new middle phase for the doubled frequency. The primary object of the present invention is to improve the deficiencies of the above mentioned conventional phase-interpolation circuits. The following description is going to disclose a novel phase-interpolation circuit which can avoid short-circuit current, provide linearly distributed multiphase clock pulse, and maintain 50% duty cycle constantly. In addition, by applying the novel phase-interpolation circuit of the present invention, we don't have to design additional duty cycle correction circuit. Moreover, we will be able to interpolate new middle phase between both the rising and falling edges of the clock pulse for the full swing signal type (digital type) phase-interpolation circuit.